Back to overview
Resolved

Degraded performance on some models due to an upstream network issue

Oct 23, 2026 at 3:43pm UTC
Affected services
deepseek-v4-flash
gemma-4-31b
glm-5.3-flash
mimo-v2.6-pro

Resolved
Oct 23, 2026 at 7:06pm UTC

Resolved: Both upstream providers have resolved their network incidents (Cloudflare's connectivity provider and our Finland infrastructure provider). Error rates and latency have been back to normal across all models since ~17:30 UTC. We apologize for the disruption.

Updated
Oct 23, 2026 at 4:02pm UTC

Update: We've identified two separate upstream issues. (1) Cloudflare is investigating degraded network performance between its Ashburn and Atlanta data centers (https://www.cloudflarestatus.com/incidents/sufgcbe8rrwt), which caused intermittent connection errors (HTTP 520/525) across all models between ~15:20 and ~15:45 UTC; error rates have since dropped back to near normal. (2) A network incident at one of our infrastructure providers in Finland is causing higher latency on mimo-v2.6-pro, glm-5.3-flash, deepseek-v4-flash and gemma-4-31b; it is improving. We'll update when both are resolved.

Updated
Oct 23, 2026 at 3:49pm UTC

Update: The same upstream network issue also caused a small number of connection errors (HTTP 520/525) across all models between ~15:20 and ~15:45 UTC. Errors have dropped sharply since; we're continuing to monitor.

Created
Oct 23, 2026 at 3:43pm UTC

We're seeing elevated latency and intermittent connection issues for some models, caused by a network incident at one of our infrastructure providers in Finland (started ~15:20 UTC). Requests are being served, but some may see slower time-to-first-token or occasional dropped connections. Other models are unaffected. We're monitoring closely and will update when the provider resolves the issue.